Machine Learning Clustering of Water-Water Interactions in the Cambridge Structural Database.

Abstract

The two machine learning methods (DBSCAN and GMM) have been applied for the analysis of water-water contacts found in crystal structures deposited in the Cambridge Structural Database. The initial dataset was divided into three groups regarding interaction energies between water molecules, and were separately analyzed. The application of machine learning methods enabled finding similar groups of contacts and defining their geometrical parameters. By carefully scrutinizing geometric parameters and visually examining clustering results, we demonstrated how valuable insights into the diverse spectrum of interactions between water molecules can be gained. Expanding the applicability of clustering methods can be achieved by integrating them into existing software for visualizing crystal structures. This approach has the potential to enhance our understanding of material behavior and accelerate the discovery of new materials.